Wednesday, March 1, 2017

The Japanese TOKI diplomatic cipher 1943-45

During WWII Japan’s Foreign Ministry used several cryptologic systems in order to protect its diplomatic communications from eavesdroppers. In 1939 the PURPLE cipher machine was introduced for the most important embassies, however not all stations had this equipment so hand ciphers continued to play an important role in the prewar period and during the war.

The main hand systems were transposed codes.

Historical overview

The first Japanese diplomatic system identified by US codebreakers was introduced during WWI and it was a simple bigram code called ‘JA’. There were two code tables, one of vowel-consonant combinations and the other of consonant-vowel. Similar systems, some with 4-letter code tables were introduced in the 1920’s.

These unenciphered codes were easy to solve simply by taking advantage of the repetitions of the codegroups of the most commonly used words and phrases. US codebreakers solved these codes and thus learned details of Japan’s foreign policy. During the Washington Naval Conference the codebreakers of Herbert Yardley’s Black Chamber  were able to solve the Japanese code and their success allowed the US diplomats to pressure the Japanese representatives to agree to a battleship ratio of 5-5-3 for USA-UK-Japan. However this success became public knowledge when in 1931 Yardley published ‘The American Black Chamber’, a summary of the codebreaking achievements of his group. The book became an international best seller and especially in Japan it led to the introduction of new, more secure cryptosystems.

In the 1930’s the Japanese Foreign Ministry upgraded the security of its communications by introducing the RED and PURPLE cipher machines and by enciphering their codes mainly with transposition systems.

Japanese transposed codes J-16 to J-19

The J-19 code had bigram and 4- letter code tables similar to the ones used previously by the Japanese Foreign Ministry. According to the NSA study ‘West Wind Clear: Cryptology and the Winds Message Controversy A Documentary History’ it was used from 21 June 1941 till 15 August 1943.

In terms of security the J-19 FUJI and the similar codes J-16 MATSU to J-18 SAKURA, that preceded it in the period 1940-41, were much more sophisticated than the older Japanese diplomatic systems. They had roughly double the number of code groups at ~1.600, these included 676 bigram entries and in addition there was a 4-letter table with 900 entries for ‘common foreign words, usually of a technical nature, proper names, geographic locations, months of the year, etc’.

These codes were enciphered mainly by columnar transposition based on a numerical key, with a stencil being used for additional security. The presence of ‘blank’ cages in the box created irregular lengths for each column of the text.

Examples of the stencils and numerical keys from ‘West Wind Clear’:

The replacements of J-19 FUJI

In summer ’43 J-19 FUJI was replaced by three new systems. The transposed codes TOKI and GEAM and the enciphered code ‘Cypher Book No1’.

TOKI was used in the period 1943-45 and it was similar to J-19 in that it was a code transposed on a stencil. The TOKI system was used by Japan’s embassies and consulates in Europe (2).

Just like its predecessor it was solved by the Anglo-Americans and the German codebreakers.

Allied exploitation of the TOKI cipher

The US effort

The TOKI transposed code was different from its predecessor J-19 FUJI in that it was used by European posts, while J-19 was used by Japanese diplomatic missions from around the world. Also TOKI was made up of 2 and 3 letter code groups while J-19 had 2 and 4 letter groups. The code groups were arranged in a non systematic manner thus making solution more difficult (3).

Examples of recovered code values (4):

The TOKI messages were enciphered using stencils and transposition keys that changed within the same message. Specifically the indicator of the message designated 3 stencils and 3 numerical keys to be used in encipherment. Each table had 250 blocks (25x10) but 50 were crossed out according to a specific system, thus 200 letters could be enciphered. If the message was longer than that then the next stencil and numerical key designated by the indicator was used.

Initially the date of the message and the signature of the originator were used to select null and blank blocks (5). In December 1943 this procedure was changed. Null blocks were abolished and the new procedure for crossing out blocks was the following (6):

at the intersection of the column and row; five blanks are inserted; the odd blanks are inserted vertically, the even blanks horizontally, for numbers 1-10 in numerical order. Blanks to be inserted below row 10 are continued at row 1. If a blank is already present in a space to be used for another blank, it is skipped over; always five blanks are inserted for each intersection point, so that the total number of blanks is 50, and the number of letters in the matrix is 200. Three such matrices and three such random sequences are used for each indicator, if the length of the message warrants this. If a message is longer than 600 textual letters, the first sequence is used for a fourth block, the second for a fifth block, etc..

Examples of stencils and transposition keys (7):

The use of a 2 and 3 letter code together with different stencils and transposition keys made solution difficult. The main method used was to analyze a large number of messages ‘in depth’ (enciphered with the same settings, identified by having the same indicator), then it was possible to use anagramming in order to solve the encipherment and recover the code values.

The report SRH-361 ‘History of the Signal Security Agency Volume Two The General Cryptanalytic Problems’, p283 says about TOKI/JBA:

JBA, a transposition system of a degree of security second only to the Purple machine-cipher system (JAA), was solved by statistical methods within six weeks. This solution is believed to be the first instance of the recovery of an unknown transposition of an unknown code by purely statistical means. Beginning groups, and later, code groups within the body of the text were found by matching stretches of cipher text from several messages with the same indicator. Frequent digraphs were recorded, and eventually the transposition patterns and tetragraphic code groups were recovered despite the presence of occasional trigraphic groups, the use of blanks in the matrix, and the use of the letters of the signature as nulls throughout the message.

Apart from statistical methods it was possible to solve messages by taking advantage of operator mistakes such as sending the same message in two different keys, enciphering the same message on TOKI and GEAM ciphers, having stereotyped beginnings etc

The Gee-Whizzer

During the war traffic on the J-series codes increased significantly and the solution of the daily changing settings became a problem for the small number of people working on Japanese systems, so there was an effort to automate the process. The device built was an attachment for standard IBM punch card equipment called the ‘Electromechanagrammer’ or ‘Gee-Whizzer’.

According to the NSA study ‘It Wasn’t All Magic: The Early Struggle to Automate Cryptanalysis, 1930s – 1960s’, p50-51:

‘The Gee Whizzer had been the first to arrive. In its initial version it did not look impressive; it was just a box containing relays and telephone system type rotary switches. But when it was wired to one of the tabulating machines, it caused amazement and pride. Although primitive and ugly, it worked and saved hundreds of hours of dreadful labor needed to penetrate an important diplomatic target. It proved so useful that a series of larger and more sophisticated "Whizzers" was constructed during the war……………….When the Japanese made one of their diplomatic "transposition" systems much more difficult to solve through hand anagramming (reshuffling columns of code until they made "sense"), the American army did not have the manpower needed to apply the traditional hand tests.

Friedman's response was to try to find a way to further automate what had become a standard approach to mechanically testing for meaningful decipherments……………………………………..Rosen and the IBM consultants realized that not much could be done about the cards; there was no other viable memory medium. But it was thought that it might be possible to eliminate all but significant results from being printed. Rosen and his men, with the permission and help of IBM, turned the idea into the first and very simple Gee Whizzer. The Whizzer's two six-point, twenty-five-position rotary switches signalled the tabulator when the old log values that were not approaching a criterion value should be dropped from its counters. Then they instructed the tabulator to start building up a new plain-language indicator value.

Simple, inexpensive, and quickly implemented, the Gee Whizzer reinforced the belief among the cryptoengineers in Washington that practical and evolutionary changes were the ones that should be given support.’

Importance of the TOKI system

From the available statistics on the solved TOKI messages and the reports issued it is clear that it was one of the high level Japanese diplomatic hand systems (together with JBB/GEAM and JBC/Cypher Book No1) (8).

In the period 1943-45 the main Japanese diplomatic systems decoded and forwarded to the Military Intelligence Service were the Purple cipher machine (JAA), the ciphers TOKI (JBA), GEAM (JBB), Cypher Book No1 (JBC) and the unenciphered code LA (JAH).

The Australian effort

In Australia the Diplomatic Special Section (D Special Section) of the Australian Military Forces HQ in Melbourne decrypted Japanese diplomatic ciphers. This unit was headed in the period 1942-44 by A.D. Trendall, Professor of Greek at Sydney University. Despite the small size of the unit considerable success was achieved in the solution of Japanese communications (9).

According to the report ‘Special Intelligence Section report - Japanese Diplomatic ciphers’ (10) the TOKI cipher was the first of the new Japanese Foreign Office ciphers to be broken.

The system was quickly compromised by the Japanese Ambassador in Lisbon Morishima Morito. The report says that he committed a fatal mistake by sending the same message in two different keys. This allowed the two messages to be solved and a few code groups to be identified.

More codegroups were recovered when some messages were sent both in the TOKI and GEAM ciphers. Since GEAM (JBB) was easier to solve it was then possible to identify the equivalent groups in solved TOKI (JBA) messages. When the cipher was modified in December 1943 it was possible to break in again by solving two messages sent in the same key.

Regarding the content of the messages the report says:

BA was used only to a moderate extent and the material it contained was of varying interest ranging from general Tokyo circulars upon international happenings to dull routine matters about couriers. Most BA messages from Russia were on the subject of couriers, visas and rations. However Stockholm was in the habit of sending all his chōhōsha (spy reports) in BA and much information was obtained therefrom. Although the second system of BA cypher might well have proved unbreakable the Foreign Ministry did not regard it very highly and issued instructions that it was to be used only for routine matters; more confidential material was to be sent in the recyphering tables. This was satisfactory from our point of view as we encountered far more difficulty in breaking and reading the second BA system than we did in recovering recyphering tables’.

The German effort

Foreign diplomatic codes and ciphers were worked on by three different German agencies, the German High Command’s deciphering department – OKW/Chi, the Foreign Ministry’s deciphering department Pers Z and the Air Ministry’s Research Department - Reichsluftfahrtministerium Forschungsamt.

OKW/Chi effort

At the High Command’s deciphering department - OKW/Chi, Japanese diplomatic systems were worked on by a subsection of Referat 13, headed by 1st Lieutenant dr Adler. About 15 people were employed by the unit (11) and according to Reinhard Wagner (a member of the section) the TOKI cipher was solved by the department.

Wagner said in his postwar interrogation report (12):

(3) A transposition procedure (Wuerfelverfahren), on which WAGNER did not work himself and which he knew only through having translated messages passed in the system. He could say of this system only that there was a daily changing keyword, and the reciphering process was complicated by Raster. The system remained valid until August 1943.

(4) The successor to the above transposition procedure, which WAGNER helped to solve, employed a basic 2 and 4 letter code book. Transposition was done in a width of 25 and a depth of 10. The keyword was changed arbitrarily. Not all the fields in the transposition square were employed but gaps (Loecher) were left. For example, the first square in the first column was to be left blank, the second square down in the second column, and so forth up to ten. In the eleventh column the top five squares down might be left blank, and in the twenty-first column the bottom five squares. In January 1944 the procedure was complicated by causing blank squares to be left vertically and horizontally. E.g., in column one, starting from the top down five squares were to be left blank. In column two, starting with the second. square down, five squares horizontally were to be left blank. In column three, starting with the third square down, five squares vertically were to be left blank, etc. The referat was successful in breaking this system.

At OKW/Chi they not only solved the Japanese transposed codes but also built a specialized cryptanalytic device called the ‘Bigram search device’ (bigramm suchgerät) for recovering the daily settings. EASI vol3, p65 says:

FUJI, a transposition by means of a transposition square with nulls applied to a two and four letter code. This system was read until it ended in August, 1943. It was broken in a very short time by the use of special apparatus designed by the research section and operated by Weber. New traffic could be read in less than two hours with the aid of this machine.

The ‘Bigram search device’ is called ‘digraph weight recorder’ in the US report ‘European Axis Signal Intelligence in World War II’ volume 2. In pages 51-53 details are given on the operation of this device:

The digraph "weight" recorder consisted of: two teleprinter tape reading heads, a relay-bank interpreter circuit, a plugboard ‘’weight’’ assignor and a recording pen and drum.
Each head read its tape photoelectrically, at a speed of 75 positions per second.

The machine could find a solution in less than two hours and did the work of 20 people, thus saving manpower.

Pers Z effort

At the Foreign Ministry’s deciphering department Pers Z Japanese systems were worked on by a group headed by Senior Specialist dr Rudolf Schauffler. This section successfully solved the Japanese diplomatic transposed codes, including the TOKI cipher which in Pers Z reports was designated as JB-64.

Dr Schroeter, a cryptanalyst of the mathematical research section who worked on Japanese ciphers, said in TICOM report I-22, p17

136. Dr. Schroeter: Had joined the organization comparatively later (Spring 1941) and had no intention of ‘staying on'. He was a lecturer in mathematical logic at the University of Münster. He had joined Dr. Kunze’s party and worked independently on Japanese recypherments.

137. He started work on simple transposition recypherments of codes; they were single transpositions with nulls over two-letter books. In the autumn of 1942-43 he worked on a Japanese 'Greater East Asia‘ traffic consisting of single transposition over a two-letter book systematically constructed, groups consisting of cv. The cage was 6 letters long and 5 or 10 letters deep with blanks evenly distributed throughout; there were three keys.

138. The system used with European posts consisted of transposition with a 25 place stencil. The stencil changed sometimes as often as three times within the message. The blanks in the stencil were filled out with the originator's signature, e.g. SHIGEMITSU. The basic book was more difficult and employed groups of two or three letters. The system was broken largely-owing to a twelve part message from Moscow with similar beginnings to each part. This system, known as 'JB 64’, is still current, though the stencil changes more frequently. Dr. Olbricht used to work on it. Dr. Schroeter sketched a specimen stencil.

It is interesting to note that a cryptanalytic device called ‘Spezialvergleicher’ was used to solve the Japanese transposed codes (13).

In the TICOM collection of the German Foreign Ministry’s Political Archive there are several folders containing worksheets of solved JB-64 messages for the period 1943-45 (14).

For example (15):

Forschungsamt effort

At the Air Ministry’s Research Department Japanese systems were worked on by Abteilung 7 (USA, UK, Ireland, South America, Spain, Portugal, Turkey, Egypt, Far East). The department had about 60-70 workers.

Unfortunately at this time there is limited information on the Forschungsamt cryptanalytic effort. In TICOM report I-25, p7 dr Martin Paetzel (deputy director of Main Department IV - Decipherment) said that a Japanese transposed code was worked on in the middle of 1943 but it was not read currently. It is possible that he was referring to TOKI.

Messages from the Japanese embassy in the Soviet Union

The Germans were particularly interested in the communications of the Japanese diplomats in the Soviet Union. It seems that this embassy was either not given a PURPLE machine or perhaps they had to dismantle it in 1941, so they relied on hand ciphers for their most important messages.

During WWII Japan fought on the side of the Axis but was careful to avoid a confrontation with the Soviet Union. War between the SU and Japan finally broke out in August 1945 but during the period 1941-45 Japanese diplomats were free to collect and transmit important information from the SU on military and political developments as well as their discussions and negotiations with Soviet officials. These messages were a prime target for the Allied and German codebreakers.

In the period 1943-45 the messages of the Japanese ambassador clearly showed the deterioration of Soviet-Japanese relations. Some of these messages were used in a series of reports prepared by Giselher Wirsing, an accomplished author and journalist, who in 1944 joined the Sicherheitsdienst foreign intelligence department as an evaluator.

Wirsing had come to the attention of General Schellenberg (head of SD foreign intelligence) due to his clear headed analysis of the global political situation and of Germany’s poor outlook for the future. Under Schellenberg’s protection he wrote a series of objective reports (called Egmont berichte) showing that Germany was losing the war and thus a political solution would have to be found to avoid total defeat (16).

In his postwar interrogations Wirsing mentioned the decoded messages of the Moscow embassy that he used in his reports:

Japanese ambassador in Moscow to his Government. Occasional telegrams were deciphered which indicated clearly that the Japanese were having increasing difficulties in maintaining friendly relations with the USSR. Through this source came confirmation from an Amt VI Far East V-man regarding a secret meeting of Japanese and Russian emissaries somewhere in SIBERIA’.

When STALIN delivered his famous address on 7 November 1944, singling JAPAN out as an aggressor nation, WIRSING, in a special report written at the request of SCHELLENBERG, read into this sentence the accomplished fact of a fundamental change of Russian policy towards JAPAN. Again SCHELLENBERG demurred. Then, approximately three weeks later, a report by ambassador SATO to his government was intercepted in which he related a conversation he had had with MOLOTOV in connection with a Japanese note expressing concern over anti-Japanese utterances by a Russian colonel in a public address. MOLOTOV, according to SATO, availed himself of this opportunity to advise the Japanese Government not to mistake rhetorical exuberance for an expression of the considered policy of the Kremlin. However, MOLOTOV added, the time would come when certain outstanding questions of a more fundamental nature would have to be thrashed out between the two nations.‘

It is reasonable to assume that some of these messages were enciphered with the TOKI system.


(2). TICOM report I-22, p17 and US report ‘The solution of the Japanese transposed code JBA’, p1 (NARA - RG 457 - Entry 9032 – NR 2828)

(3). US report ‘The solution of the Japanese transposed code JBA’ (NARA - RG 457 - Entry 9032 – NR 2828)

(4). US report ‘Master JBA trigraph charts‘ (NARA - RG 457 - Entry 9032 – NR 2458)

(5). The procedure as described in ‘The solution of the Japanese transposed code JBA’ was as follows. The cipher clerk would take the stencil, write the numerical key at the top and then add the letters of the signature of the originator (for example SATOAMBASS) at the blocks on row 1-column 1, row 2-column 2, row 3-column 3 etc.

The blank blocks were selected from a key table which identified the columns to be crossed out for each day of the month.

(6). US report ‘The solution of the Japanese transposed code JBA’ (NARA - RG 457 - Entry 9032 – NR 2828)

(7). US reports ‘Report on Japanese diplomatic systems 1944’ (NARA - RG 457 - Entry 9032 – NR 3095) and ‘The solution of the Japanese transposed code JBA’ (NARA - RG 457 - Entry 9032 – NR 2828)

(8). US report 'Foreign Cryptographic Systems, 1942-1945' (NARA - RG 457 - Entry 9032 - NR3254)

(9). Breaking Japanese Diplomatic Codes David Sissons and D Special Section during the Second World War, chapter 1

(10). Australian National Archives - NAA: A6923, 1/REFERENCE COPY - (barcode 12127133)

(11). TICOM I-124, p3 and TICOM I-150, p4

(12). TICOM I-90 ‘Interrogation of Herr Reinhard Wagner (OKW/Chi) on Japanese systems’, p3

(13). TICOM I-22, p18

(14). German Foreign Ministry’s Political Archive - TICOM collection - files Nr. 2.465-2.471

(15). German Foreign Ministry’s Political Archive - TICOM collection - file Nr. 2.471 - Japan 1944/45   ‘JB64, 1481-1644’, Diplom. Briefverkehr

(16). British national archives KV 2/140 ‘Giselher WIRSING: Journalist and author’


‘European Axis Signal Intelligence in World War II’ volumes 2,3,6,7 , TICOM reports I-22, I-25, I-90, I-124, I-150,  DF-187B , ‘The Codebreakers’, ‘Breaking Japanese Diplomatic Codes David Sissons and D Special Section during the Second World War’, United States Cryptologic History Series IV: World War II Volume X: ‘West Wind Clear: Cryptology and the Winds Message Controversy A Documentary History’, United States Cryptologic History, Special Series, Volume 6, ‘It Wasn’t All Magic: The Early Struggle to Automate Cryptanalysis, 1930s – 1960s’, NSA interviews of Frank Rowlett 1974 (National Cryptologic Museum Library), Australian National archives: ‘Special Intelligence Section report - Japanese Diplomatic ciphers’, NARA reports: ‘The solution of the Japanese transposed code JBA’, 'Foreign Cryptographic Systems, 1942-1945', ‘Report on Japanese diplomatic systems 1944’, ‘Master JBA trigraph charts‘, NSA report SRH-361 ‘History of the Signal Security Agency Volume Two The General Cryptanalytic Problems

Acknowledgements: I have to thank Rene Stein for identifying several of the NARA JBA reports.


  1. Very nice research Chris, but I have spotted one error probably due to a confusion about the terms. You say:
    They had roughly double the number of code groups at ~1.600 bigram entries and in addition there was a 4-letter table with 900 entries for ‘common foreign words, usually of a technical nature, proper names, geographic locations, months of the year, etc’

    The number ~1.600 bigram entries is wrong. A bigram code can only have 676 entries (26 x 26). The figure 1600 refers to the total number of code groups, 676 bigram codes and 900 4-letter, which gives 1576, your ~1.600 entries. I am recently doing some research into early (1939-1942) Japanese diplomatic codes and ciphers and I therefore find your research very interesting.

    1. Fixed.

      I agree that the FUJI and TOKI ciphers were very interesting. Quite a lot of work went into this essay.

      Glad to hear you’re working on new non-Enigma articles.